This is a convolutional neural network trained to distinguish sober headlines from more clickbaity ones. It's an implementation of Yoon Kim's 2014 paper, Convolutional Neural Networks for Sentence Classification. It is trained using a dataset pulled from reddit's news subreddits, several RSS feeds from Reuters, BBC, and CBC, while the clickbait comes from Buzzfeed and Viralnova. The word vectors used are pretrained, namely from GloVe. It is implementated using Tensorflow, with a web app made in Flask which communicates with a Tensorflow Serving server.
The model gets 95.95% accuracy and a F1 score of 0.9705 after 5 epochs.
There is a public demo of the classifier here
